CN115551013A - Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network - Google Patents
Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network Download PDFInfo
- Publication number
- CN115551013A CN115551013A CN202211106470.7A CN202211106470A CN115551013A CN 115551013 A CN115551013 A CN 115551013A CN 202211106470 A CN202211106470 A CN 202211106470A CN 115551013 A CN115551013 A CN 115551013A
- Authority
- CN
- China
- Prior art keywords
- task
- unmanned aerial
- aerial vehicle
- ground terminal
- follows
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0925—Management thereof using policies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/18502—Airborne stations
- H04B7/18506—Communications with or from aircraft, i.e. aeronautical mobile service
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
Abstract
The invention provides a method for deploying and unloading tasks of unmanned aerial vehicles in a multi-unmanned aerial vehicle edge computing network, aiming at maximizing the number of the tasks borne by the unmanned aerial vehicles. Mainly comprises the following steps: 1. and constructing a mathematical model P1 for unmanned aerial vehicle deployment and task unloading in the multi-unmanned aerial vehicle edge computing network. 2. Under the condition that the calculation frequency of a CPU (central processing unit) allocated to the task by the given unmanned aerial vehicle is given, a mathematical model P2 is constructed, the problem P2 is solved based on a differential evolution algorithm and a greedy algorithm, and the optimal position and elevation angle of the unmanned aerial vehicle, the transmission power of ground terminal equipment and the unloading decision of the task are solved. 3. And constructing a mathematical model P3 based on the obtained optimal position and elevation angle of the unmanned aerial vehicle and the transmission power of the ground terminal equipment, and obtaining the CPU calculation frequency and unloading decision of the unmanned aerial vehicle distributed to tasks. 4. And (5) circularly and iteratively solving between the step 2 and the step 3. 5. The application improves the task execution efficiency of the unmanned aerial vehicle in the multi-unmanned aerial vehicle edge computing network.
Description
Technical Field
The invention belongs to the technical field of wireless networks, and relates to a method for deploying unmanned aerial vehicles and unloading tasks in a multi-unmanned aerial vehicle edge computing network.
Background
With the development of the internet of things, wireless communication devices and wireless data traffic have increased geometrically. Especially in recent years, with the emergence of various computationally intensive and delay sensitive applications such as multimedia video streaming services, augmented reality, virtual reality, intelligent transportation, etc., it is important to provide a network with low delay, ultra-reliability, and high robustness. Cloud computing in a traditional core network is difficult to meet the requirements of service quality and service experience of terminal equipment due to long service delay and serious network congestion. Moving Edge Calculation (MEC) arises. The mobile edge computing is a network architecture which can realize cloud computing service on the edge side of the network, and can reduce time delay, improve efficiency and save power consumption, thereby improving service experience of users. But the deployment of MECs also has not minor drawbacks. First, MECs are expensive to deploy for hot spot areas of a city, such as commercial districts or densely populated areas. The network coverage is limited for rural areas, mountain areas, remote areas such as the sea and the like and areas which are easily affected by natural disasters, and edge unloading service is difficult to provide.
In recent years, unmanned aerial vehicles have been rapidly developed, and unmanned aerial vehicle edge computing provides a new solution for solving the above problems. The unmanned aerial vehicle edge computing refers to combining an edge computing architecture with an unmanned aerial vehicle platform, and the unmanned aerial vehicle can be used as a user node to unload a computing-intensive task to an edge server located on a ground base station, and can also be used as an aerial edge server to provide computing unloading service for a plurality of ground user nodes. Compared with ground communication, the unmanned aerial vehicle has the characteristic of flexibility, can provide anytime and anywhere mobile computing service even in mountainous areas, oceans and complex terrains, and provides spatial freedom. Meanwhile, the unmanned aerial vehicle can establish an air-ground line-of-sight link with the ground terminal equipment with a high probability, so that the quality of a communication link is improved, and the calculation performance is improved.
When the ground terminal equipment is numerous and the ground terminal equipment is required to execute tasks with large task amount and high calculation requirement, the limited calculation resource and coverage range of a single unmanned aerial vehicle can not meet the requirement of the ground terminal equipment, and multiple unmanned aerial vehicles can meet the requirement of the ground terminal equipment well. Therefore, when ground terminal equipment needs to execute huge calculation tasks, the unmanned aerial vehicle deployment decision-making method and the unmanned aerial vehicle resource allocation method aim at maximizing the number of the bearing tasks of the unmanned aerial vehicle according to the position of the terminal equipment, the channel state, the available CPU calculation resources of the unmanned aerial vehicle and other information.
In view of the above considerations, the present invention provides a method for unmanned aerial vehicle deployment and task offloading in a multi-unmanned aerial vehicle edge computing network.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method for deploying and unloading tasks of unmanned aerial vehicles in a multi-unmanned aerial vehicle edge computing network. By jointly optimizing the task offloading decisions, resource allocation, and the position and elevation of the drone, the goal is to maximize the number of load-bearing tasks. The technical scheme of the invention is as follows:
a method for deploying unmanned aerial vehicles and unloading tasks in an edge computing network of multiple unmanned aerial vehicles comprises the steps of firstly constructing an unmanned aerial vehicle-assisted mobile edge computing system, wherein the system consists of M ground terminal devices and N unmanned aerial vehicles carrying edge servers, psi = {1,2, · i,.., M } represents a set of the ground terminal devices, phi = {1,2,.., j,..., N } represents a set of the unmanned aerial vehicles, and all the unmanned aerial vehicles are provided with directional antennas with adjustable beam widths. Assuming that each ground terminal device has 1 task at the current moment, the calculation task of the ground terminal device i e ψ can be executed locally or unloaded to the unmanned aerial vehicle for execution.
The invention provides a method for deploying unmanned aerial vehicles and unloading tasks in a multi-unmanned aerial vehicle edge computing network, which comprises the following steps: 1. a mathematical model of unmanned aerial vehicle deployment and task unloading in a multi-unmanned aerial vehicle edge computing network is established, and the method comprises the following steps:
the Task of each ground terminal device i epsilon psi is expressed as a binary Task i =(C i ,D i ) In which C is i Representing processing tasks Task i The required CPU periodicity with the unit of cycles; d i Task representation i The unit is bit. (x) i ,y i ) Indicating the location, x, of the ground terminal equipment i e ψ i And y i The plane coordinates of the ground terminal equipment are all expressed in m, and the height is all 0 meter (m).
The 3-dimensional coordinates of each drone j ∈ φ are represented as (X) j ,Y j ,H j ) Wherein X is j And Y j For the plane coordinate of unmanned plane j ∈ phi, H j Height, in m, representing drone j ∈ φ. And the unmanned plane j epsilon phi is provided with a directional antenna with adjustable beam width. The azimuth angle and elevation angle half-power beam widths of the unmanned aerial vehicle j epsilon phi antenna are equal and are all 2 theta j And e (0, pi) represents. Unmanned plane j belongs to phi position and elevation angle Q j =(X j ,Y j ,H j ,θ j ) The position and elevation of all drones is denoted by Q.
a i,j Task for Task i Unload decision of a i,j =0 for local calculation, a i,j =1 denotes Task to be executed i And unloading to unmanned plane j for calculation. The unloading of the ground terminal device task comprises two stages: 1) Ground terminal equipment i epsilon psi sends Task i To drone j, 2) drone j allocates a computing resource computing task. Each ground terminal device is fixed in position before the mission is completed.
Task of S1-1 ground terminal equipment i epsilon psi i Execute locally
Task of ground terminal equipment i belonging to psi i The time of execution in the local is shown as (1):
whereinAnd the CPU frequency (in cycles/s) of the local processing task of the ground terminal equipment i epsilon psi is shown.
Ground terminal equipment i belongs to psi and executes Task i The energy consumption calculated locally is shown in (2):
wherein eta l The effective capacitance coefficient represented by the ground terminal device i ∈ ψ depends on the chip architecture of the CPU.
S1-2 Task i Offloading to unmanned aerial vehicle execution
S1-2-1 communication model
The plane distance between the ground terminal equipment i epsilon psi and the unmanned plane j epsilon phi is shown as (3):
unmanned plane j 1 With unmanned plane j 2 The distance between the planes is shown as (4):
unmanned plane j 1 With unmanned plane j 2 The distance between them is shown as (5):
the azimuth angle half-power wave beam width and the elevation angle half-power wave beam width of the unmanned plane j epsilon phi antenna are equal, and the antenna gain modeling in the azimuth angle theta direction and the elevation angle u direction is shown in (6):
wherein G is 0 2.2846, ω =0 represents the channel gain outside the antenna beam width. For simplicity, ω =0 is set.
The user i belongs to psi and unloads the task to the unmanned plane j belongs to phi, and then the user i belongs to psiMust be within the coverage of drone j e phi, as shown in (7)
Channel gain g between ground terminal equipment i epsilon phi and unmanned aerial vehicle j epsilon phi i,j As shown in (8):
wherein beta is 0 Representing the channel power gain at a reference distance of 1 m.
The uplink transmission rate from the ground terminal device i e psi to the drone j e phi is shown as (9):
the uplink transmission power from the ground terminal device i epsilon psi to the drone j epsilon phi is shown as (10):
wherein p is i,j Representing the transmission power (in W) from user i e ψ to drone j e. B represents the channel bandwidth between the ground terminal equipment i epsilon psi and the unmanned plane j epsilon phi, and the bandwidth between all the ground terminal equipment and the unmanned plane is equal and has the unit of Hz. N is a radical of 0 Representing the noise power spectral density (in W/Hz).
The transmission time from the ground terminal equipment i epsilon phi to the unmanned aerial vehicle j epsilon phi for unloading the task is shown as (11):
the transmission energy from the ground terminal equipment i epsilon psi to unload the task to the unmanned aerial vehicle j epsilon phi is shown as (12):
s1-2-2 computational model
Task i The time to unload to drone j e phi execution is shown as (13):
whereinIndicates that the unmanned plane j belongs to phi and is allocated to the Task i The CPU frequency (unit is cycles/s).
The energy consumption of the calculation executed by the task offloaded to the unmanned plane j ∈ phi is shown as (14):
wherein mu c Representing the effective capacitance coefficient of the drone.
S1-3 problem description
Defining a mathematical model P1, and under the condition of meeting constraint conditions, maximizing the number of the bearing tasks of the unmanned aerial vehicle, wherein the model is as follows:
equation (15 a) is an objective function, χ = (F, a, Q, P) represents an optimization variable, whereRepresenting the CPU frequency allocation of the drone,the decision to offload a task is made, indicating the position and elevation of the drone,representing the transmission power of the ground terminal equipment.
Equation (15 b) represents a completion time constraint for the task, both executed locally or on the drone, less than T max 。
Equation (15 c) represents a maximum energy consumption constraint for executing the task, with energy consumption less than E for executing locally or on the drone max 。
Equations (15 d) and (15 e) represent the maximum CPU computation frequency constraints of the ground terminal device and the drone, whereIs the maximum CPU computation frequency for the local computation,is the CPU computation frequency of a single drone.
Equation (15 f) represents the unloading decision constraint of the ground terminal equipment, and each task can be only unloaded to one unmanned aerial vehicle.
Equation (15 g) represents the maximum transmission power constraint for uplink transmission in the system, whereMaximum transmission power when the task is uploaded.
Equation (15 h)) represents the coverage constraint of the drone to which a ground terminal device within the coverage of the drone can offload tasks.
Equation (15 i) represents the constraint of minimum distance between drones.
Equation (15 j) represents the constraint on the altitude of the drone.
Equation (15 k) represents the constraint on the elevation (half-power beamwidth) of the drone.
Expression (15 l) indicates that the coverage areas of the drones do not overlap.
P1 is a non-convex non-linear optimization problem, and therefore, the conventional optimization method cannot solve the problem. P1 is divided into 2 sub-problems, the first of which is the assignment of a given drone to a Task i The CPU calculates the frequency F, solves the position and elevation angle Q of the unmanned aerial vehicle, the transmission power P of the ground terminal equipment and the unloading decision A of the unmanned aerial vehicle. The second sub-problem is to solve the final unloading decision a of the drone and the assignment of the drone to the Task given the location and elevation Q of the drone and the transmission power P of the ground terminal equipment i The CPU of (1) calculates the frequency F. Let δ =1, δ denote the number of iterations of the whole, δ max Represents the maximum iteration number, and O (delta) represents the number of tasks carried by the unmanned aerial vehicle after delta iterations
2. The steps of task unloading decision, optimal position and elevation angle of the unmanned aerial vehicle and transmission power of the ground terminal equipment based on a differential evolution algorithm and a greedy algorithm are as follows:
s2-1, constructing a mathematical model under the condition that the CPU of the unmanned aerial vehicle is assigned to each task to calculate the frequency F. And under the condition that the CPU calculation frequency F distributed to each task by the unmanned aerial vehicle is given, the unloading decision of the task, the optimal position and the elevation angle of the unmanned aerial vehicle and the transmission power of the ground terminal equipment are optimized, and a differential evolution algorithm and a greedy algorithm are adopted for solving. And solving the problem that the target is consistent with the problem P1, and maximizing the number of the bearing tasks of the unmanned aerial vehicle. The mathematical model P2 of the optimization problem can be written as follows:
(16a) For the objective function, (16 b) - (16 k) are constraints, Ω = (a, Q, P) where Are the optimization variables.
S2-2, adopting a differential evolution algorithm and a greedy algorithm to solve unloading decision of tasks, optimal position and elevation angle of the unmanned aerial vehicle and transmission power of ground terminal equipment, and specifically comprising the following steps:
solving a problem P2, under the condition that the CPU calculation frequency F distributed to each task by the unmanned aerial vehicle is given, firstly, solving the position and the elevation angle Q of the unmanned aerial vehicle based on a differential evolution algorithm, solving the transmission power P of the ground terminal equipment and the unloading decision A of the task based on a greedy algorithm, and finally obtaining the maximum bearing task number of the unmanned aerial vehicleSetting maximum iteration times tau of differential evolution algorithm max =10000, and let τ =1 be the current iteration number.
The steps of solving the problem P2 by the differential evolution algorithm are as follows:
i) The population is initialized as follows:
a coding mechanism is proposed in which a population set representsWhereinRepresenting the set of position and elevation of the Tth iteration drone, Q j (τ)=(X j (τ),Y j (τ),H j (τ),θ j (τ)) represents the position and elevation coding for drone j ∈ φ, while noting Q j (τ)=q j (X j (τ),Y j (τ),H j (τ),θ j (τ)) representing the position and elevation angle of the drone j ∈ Φ at the τ th iteration, and the value range is as follows:
wherein c is 1 And 0 each represents X j Upper and lower bounds of c 1 Representing a constant in meters (m). c. C 2 And 0 each represents Y j Upper and lower bounds of c 2 Denotes a constant in meters (m), H max And H min Respectively represent H j Upper and lower bounds in meters (m), θ max And theta min Respectively represent theta j An upper bound and a lower bound.
For convenience of description of X j (τ),Y j (τ),H j (τ),θ j (τ) value range, let us assumeThe s decision variable representing the unmanned plane j belongs to phi, the value of s is {1,2,3,4 }, which respectively corresponds to Q j X in (τ) j (τ),Y j (τ),H j (τ),θ j (τ) that is
(17a) - (17 d) can be collectively expressed as follows:
Randomly generating a decision variable X j (τ),Y j (τ),H j (τ),θ j The initial population of (τ) is calculated as follows:
where rand (0, 1) denotes a random selection of a constant from 0-1.
The specific process of initializing the population comprises the following steps:
(1) let j =1,j denote the serial number of the drone, and randomly generate the position and elevation Q of drone j based on equation (20) j (τ)=(X j (τ),Y j (τ),H j (τ),θ j (τ)), and Q is added j (τ) put into the set of drone positions and elevations Q (τ) of the τ th iteration.
(2) Let iota =0, iota denote the generation of the j position and elevation Q of the drone j Number of formation of (. Tau.),. Iota max =200 denotes randomly generating drone j position and elevation Q j (τ) maximum number of generations.
(3) Randomly generating position and elevation angle Q of unmanned aerial vehicle j +1 j+1 (τ), iota = iota +1. Judging whether the position of each unmanned aerial vehicle in the unmanned aerial vehicle j +1 and the unmanned aerial vehicle position and elevation angle set Q (tau) of the tau iteration meets constraint expressions (16 g) and (16 j), namely Q (tau) and Q j+1 Whether or not the position of (τ) satisfies the constraint equation (16 g) and the equation (16 j). The judgment and selection are as follows:
if Q (τ) and Q j+1 (τ) satisfies the constraint equations (16 g) and (16 j), which indicates Q (τ) and Q j+1 (τ) distance between drones is greater than or equal to the nearest distance limit of dronesI.e. no interaction between drones in drones j +1 and Q (τ)Collision, and the coverage of drones j +1 and Q (τ) do not intersect, then let j = j +1, let Q be j (tau) putting the unmanned aerial vehicle position and elevation angle set Q (tau), and skipping to the step (4). If Q (τ) and Q j+1 (τ) does not satisfy constraint equation (16 f) and equation (16 j), indicating that the generation of the j +1 st drone position was unsuccessful. Iota is less than or equal to max And (4) if the answer is positive, skipping to the step (3) if the answer is positive, otherwise skipping to the step (1).
(4) And (4) judging whether j is less than or equal to N, and if so, returning to the step (2). Otherwise, the positions and the elevation angles of all the drones are successfully generated, and the initial population Q (tau) of the drones is obtained.
ii) calculating the transmission power of the ground terminal equipment according to the new position and elevation angle of the unmanned aerial vehicle, and the steps are as follows:
and according to the position and the elevation angle of the unmanned aerial vehicle j epsilon phi and the position information of the user i epsilon psi, finding out the coverage range of which unmanned aerial vehicle the ground terminal equipment i epsilon psi is in through an equation (16 f). Calculating the frequency F and the maximum time delay T of the CPU assigned to a task for a given drone according to equation (16 b) max And obtaining the transmission rate required by the ground terminal equipment i epsilon psi to the unmanned aerial vehicle, wherein the transmission rate meets the following constraint:
as can be seen from the analysis formula (21), the transmission power of the ground terminal equipment must satisfy the formula (21), and meanwhile, the transmission power of the ground terminal equipment must also satisfy the formula (16 c), so that the equation (21) takes the equal sign as the optimal value of the transmission power of the ground terminal equipment.
Based on (21) and (10), the transmission power p of the tau iteration terrestrial terminal equipment i epsilon psi can be calculated i,j (τ) is represented by formula (22):
iii) Calculating the unloading decision of the task according to the transmission power of the ground terminal equipment and the new position and elevation angle of the unmanned aerial vehicle, and comprising the following steps:
dividing the task set psi into psi 1 、ψ 2 、ψ 3 Three categories, and satisfy ψ = ψ 1 ∪ψ 2 ∪ψ 3 。
Wherein the first type task psi 1 Having the highest priority, ground terminal equipment i epsilon psi 1 Indicating that it is not within the coverage of the drone, task at this time i Can only be done locally, in which case the unload decision a for the # th iteration i,j (τ)=0。
Task psi of the second type 2 At this time, the ground terminal equipment i is epsilon psi 2 In the coverage of unmanned aerial vehicle, task i The method can be unloaded to an unmanned aerial vehicle for execution, and the tau-th iteration ground terminal equipment i epsilon psi can be calculated based on the transmission power obtained in the step ii) 2 Transmission rate r to drone j i,j (τ). Unmanned aerial vehicle assignment to Task i Under the condition that the calculation resources satisfy the formulas (16 b) and (16 c), the unloading decision is solved based on a greedy algorithm, and the selection strategy of the greedy algorithm is that the ground terminal equipment i belongs to psi 2 Transmission rate r to drone j i,j (τ), i.e. transmission rate r i,j (τ) fastest is selected first, with the unload decision a of the τ th iteration i,j (τ)=1。
Task type III 3 That is, after the unmanned aerial vehicle computing resources are fully loaded, the tasks can only be performed locally, namely, the total task psi is divided by psi 1 、ψ 2 Task outside, unload decision a for the τ th iteration at this time i,j (τ)=0。
iv) calculating a target value
Obtaining j epsilon phi position and elevation angle of unmanned aerial vehicle of the tau iteration based on the steps ii) and iii)The transmission power of the ground terminal equipment and the unloading decision of the task can obtain the result { P (tau), Q (tau), A (tau) } of the tau-th iteration, wherein P (tau) represents the transmission power of the ground terminal equipment of the tau-th iteration, and Q (tau) represents the position set of the tau-th iteration unmanned aerial vehicleIn sum, a (τ) represents the task unload decision for the τ -th iteration. Calculating a target value of the tau iteration based on the result of the tau iteration, namely the maximum number of bearing tasks of the unmanned aerial vehicle
v) after determining the position and the elevation angle of the tau-th iteration unmanned aerial vehicle, performing cross mutation operation, wherein the steps are as follows:
mutation operation: according toIs provided with For the j epsilon phi position and elevation angle of the unmanned aerial vehicle obtained after the tau-th iterative variation, respectively correspond toBy mutation operations ofThe calculation formula of (a) is as follows:
equation (23) represents the position and elevation q for the tau-th iteration for drone j ∈ phi j (X j (τ),Y j (τ),H j (τ),θ j (τ)) performing a mutation operation, wherein j 1 、j 2 、j 3 E phi represents that 3 different numbers are randomly selected in the unmanned aerial vehicleτ denotes Q in a differential evolution algorithm j F denotes the scaling factor, taking F =0.9.
And (3) cross operation: is also provided withRepresents the j ∈ phi position and elevation angle of the unmanned plane obtained after the tau-th iteration crossing according toAnd with Is selected by cross probability to obtainWhereinThe calculation formula of (c) is as follows:
equation (24) represents the operation of crossing the tau-th iteration for drone j e phi position and elevation, CR represents the probability of crossing, taking CR =0.9.
Through mutation and crossing to obtainFinally obtaining a population U (tau) = { U = [) 1 (τ)、U 2 (τ)、…U j (τ)、…U N (τ) }. Steps ii) -iii) determining a new { P } * (τ)、Q * (τ)、A * (τ) }, step iv) obtaining the target value L * (τ)。
vi) the selection operation is as follows:
selection based on greedy thought, i.e.The largest of the target values is selected as the new individual. Crossing the variation to obtain a population U (tau) = { U = [ (U) ] 1 (τ)、U 2 (τ)、…U j (τ)、…U N (τ) } and the original population Q (τ) = { Q 1 (τ)、Q 2 (τ)、…Q j (τ)、…Q N (τ) }, the comparison formula is shown as formula (25):
comparison L * (τ) and L (τ), the steps are as follows:
if L ≧ L (τ) indicates that the population obtained after the cross mutation operation is better, { P (τ + 1), Q (τ + 1), a (τ + 1) } = { P * (τ)、Q * (τ)、A * (τ) }, if L * (τ) < L (τ), then { P (τ + 1), Q (τ + 1), a (τ + 1) } = { P (τ), Q (τ), a (τ) } is unchanged.
vii) judgment of τ +1 > τ max And if yes, stopping loop iteration, outputting { P (tau + 1), Q (tau + 1) and A (tau + 1) }, otherwise, tau = tau +1, and jumping to the step v).
3. According to the position and elevation angle Q of the unmanned aerial vehicle obtained in the step 2 and the transmission power P of the ground terminal equipment, solving a final unloading decision A of the unmanned aerial vehicle and a Task allocated to the unmanned aerial vehicle i The CPU calculates the frequency F, and the solving steps are as follows:
s3-1, solving an unloading decision and calculating resource allocation of the unmanned aerial vehicle, wherein the aim is to maximize the number of bearing tasks of the unmanned aerial vehicle, and a mathematical model P3 of an optimization problem can be written in the following form:
wherein x 1 = (A, F), whereinThe following equations (8), (9), (11) and (12) showThe portion is a known value, i.e., a constant.
And S3-2, after the position and the elevation angle Q of the unmanned aerial vehicle and the transmission power P of the ground terminal equipment are obtained according to the step 2, solving an unloading decision A of the task and a calculation resource distribution F of the unmanned aerial vehicle by adopting a convex optimization method.
The specific operation of S3-2-1 non-convex-to-convex is as follows:
i) Order toAndexpressing expressions (26 b) and (26 c), respectively, and obtaining hessian matrices of (26 b) and (26 c) as follows:
as can be seen from the analysis, since both equations (27 a) and (27 b) are non-positive definite matrices, (26 b) and (26 c) are non-convex constraints.
ii) carrying out a non-convex-to-convex transformation operation by a linearization constraint factor, comprising the following steps:
the equation (26 b) is first linearized, and a small variable ε is introduced in order to avoid the denominator being zero 1 Equation (26 b) can be restated as:
an auxiliary variable gamma is defined i,j The method comprises the following steps:
substituting formula (29) into formula (28) to give formula (30):
substituting formula (29) into formula (26 e) to give formula (31):
since equation (26 e) represents the maximum CPU calculation frequency of the drone, it can be obtainedAlso, because of the formula (29), can obtainAnd replaceIn (1)The formula (32) can be obtained finally:
then model P3 may be rewritten to model P4 as follows:
since (33 b) and (33 c) are also non-convex, the equations (33 b) and (33 c) are relaxed as follows:
(1) for the relaxation operation of equation (33 b), the procedure is as follows:
due to the offload decision a i,j E (0, 1) is a discrete variable, thus relaxing a i,j E (0, 1), as shown in (34):
for formula (34), defineFrom the equations (32) and (34), the constraint factor product ω can be obtained i,j :
(2) for the relaxation operation of formula (33 c), the procedure is as follows:
definition ofBecause ofTherefore, it is not only easy to useThen σ can be obtained i,j The ranges of (a) and (b) are as follows:
model P4 can be transformed into P5 as follows:
(38),(39) (40g)
Because of the fact thatTherefore, it is possible toAndthe trend of (1) is the same, so the target formula is changed toA mathematical model P6 was obtained as follows:
S.t. (40b),(40c),(40d),(40e),(40f),(40g) (41b)
s3-2-2: adopting a KKT condition to solve the calculation frequency of the unloading decision and the task of the unmanned aerial vehicle, and specifically comprising the following steps:
i) Solving the problem P6, constructing the Lagrangian function and dual problem of P6, and orderingWherein A is 1 Represents the task unloading decision of the model P6 through convex optimization,lagrange multipliers of expressions (40 b) - (40 c),the Lagrangian multiplier of expression (40 d),a Lagrangian multiplier represented by the formula (40 e),a Lagrangian multiplier of expression (40 f),the lagrange multipliers respectively expressed by the formula (40 g),lagrange multipliers respectively representing formula (40 j), then
The lagrangian function of problem P6 is as follows:
ii) Lagrangian function (42) for a i,j ,Calculating a partial derivative, wherein the steps are as follows:
lagrangian function (42) for a i,j The partial derivatives are calculated as follows:
iii) The conditions according to KKT are as follows:
from equation (43), equation (50) can be derived, as follows:
formula (51) is obtainable from formula (44):
formula (52) is obtained according to formula (45):
equations (53) and (54) are obtained from equations (46) and (47):
finding gamma in equation (36) i,j 、ω i,j While the related equation can be found:
therefore, it can be analyzed from the formulas (55) and (56) that (53) and (54) are γ, respectively i,j Upper and lower bounds of (a) if gamma i,j When the value between (52) and (53) does not take the boundary, thenAnd withAre both 0.
Equations (57) and (58) are obtained from equations (48) and (49):
therefore, the (57) and (58) can be analyzed according to the formulas (59) and (60)Other is σ i,j Upper and lower bound of (a), if i,j In the case that the value between (57) and (58) does not take the boundary, thenAnd withAre both 0.
By analyzing (40 d), (40 e), (40 f) and (40 g), the results were obtained Since the portion is not 0, according to the KKT condition, the section is 0. While not taking sigma i,j And gamma i,j The boundary of (2) is thenAll are 0, initialization is performed so that i =1.
Equation (50) can be simplified to (61):
iv) initializationRespectively corresponding to convergence accuracy xi 1 、ξ 2 ρ is the number of iterations, let ρ =0, and the initial value is
v) obtaining a obtained from the rho-th iteration according to the formulas (51), (52) and (61) i,j (ρ)、ω i,j (ρ)、At this time The cases of (c) are as follows:
vi) updating Lagrange multiplier based on gradient descent method according to (44) and (45)As follows:
vii) judgment of a i,j (ρ)-a i,j (ρ -1) > 0, if yes, step v) is skipped, otherwise step viii) is skipped.
viii) judging whether i > M is satisfied, if not, i = i +1, turning to step iv), otherwise, outputting A 1 、ω i,j 、
ix) calculating assignment of Task unmanned aerial vehicle j to Task i The calculation frequency of (c):
the Task can be obtained by the equations (29) and (48) i Is calculated frequency ofThe formula is as follows:
x) by A 1 The offloading decision a is obtained as follows:
wherein A (a) i,j ) Representing a in the set of offload decisions A i,j ,A 1 (a i,j ) Representing offload decision set A 1 A in i,j 。
by A 1 Reselecting an unloading decision A, starting to select j =1 from a first unmanned aerial vehicle, and specifically comprising the following steps:
i) Reselecting the unloading decision A through the residual calculation resources of the unmanned aerial vehicle, and comprising the following steps:
(1) selection A 1 0 in (A) 1 (a i,j ) The value in the interval < 0.5, and secondly recording a in this interval i,j Task corresponding to the minimum value ofIs recorded as i 1 . Let A (a) i,j )=1,A 1 (a i,j ) =0. Is newly paired by the formula (51)And (6) performing operation.
(2) If it isThen theJudgment of 0 < A 1 (a i,j ) If < 0.5 is an empty set, the empty set jumps to step ii). Otherwise, returning to the step i).
first, A is selected 1 A in the total is more than or equal to 0.5 1 (a i,j ) The value in the interval less than or equal to 1 is recorded i,j Task of maximum value ofRecord asOrder toIs re-paired by the formula (51)Carry out the operation ifThen theIf it isAnd returning to the step i) again.
iii) Let j = j +1, when j > N, then obtain a final unloading decision a, and the CPU computation frequency allocated to the upload task by the drone is F.
iv) calculatingAnd O (delta) represents the number of the tasks carried by the unmanned aerial vehicle after delta iterations, and the values of the results F, A, Q and P of the optimization variables are obtained according to P2 and the formula (67).
4. Comparing O (δ) with O (δ + 1), if O (δ 0+ 1) > O (δ 1), O (δ + 1) = O (δ + 1), and recording F, a, Q, P for O (δ + 1), otherwise O (δ + 1) = O (δ), maintaining F, a, Q, P for O (δ). Judging that delta +1 is larger than delta max And if not, making delta = delta +1, skipping to the step 2 for a new iteration based on the calculated F, and if so, ending the iteration to obtain F, A, Q and P of the optimized variables and the maximum bearing task number O (delta) of the unmanned aerial vehicle.
Advantageous effects
The invention provides a method for deploying unmanned aerial vehicles and unloading tasks in a multi-unmanned aerial vehicle edge computing network. Effectively obtain unmanned aerial vehicle's maximize and bear the weight of the task number.
Drawings
FIG. 1 is a schematic view of a scene model of the present invention;
fig. 2 is a flowchart of a method for deploying unmanned aerial vehicles and unloading tasks in a multi-unmanned aerial vehicle edge computing network according to the present invention;
FIG. 3 is a flow chart of the present invention for solving for transmission power, UAV position, and elevation;
FIG. 4 is a flow chart of the present invention for offloading decisions to solve a task and for unmanned aerial vehicle CPU calculation frequency;
Detailed Description
The invention will be described in further detail below with reference to the following figures and specific examples:
example 1:
l1 in this embodiment, fig. 1 is a schematic view of an unmanned aerial vehicle edge computing scene model, which includes N =4 unmanned aerial vehicles, is equipped with an edge server, has M =15 ground terminal devices, and has 1 task on each user. The Task of the ground terminal equipment is described as Task i =(C i ,D i ). The maximum CPU frequency of the ground terminal equipment isThe maximum CPU frequency of the edge server isSetting the maximum completion time T of each task max =1s, maximum energy consumption E max =1J. Let the transmission bandwidth be B =1MHZ, noise power N 0 =10 -20W /Hz, channel power gain at reference distance β =1.42 × 10 -4 ,G 0 =2.2846 is a normal number, maximum transmission power P max =1.5w. Minimum distance between unmanned aerial vehiclesEffective capacitance coefficient eta of unmanned aerial vehicle and terminal equipment 1 =10 -27 ,η 2 =10 -28 。
L1-1 initializes the location of the ground terminal and its task attributes, and the CPU computation frequency assigned to the task. Wherein position x i 、y i Task i C of (A) i And D i The value ranges of (a) are shown in table 1, the value ranges of the parameters of the unmanned aerial vehicle are shown in table 2, the attribute information of the task is shown in table 3, the calculation frequency initially allocated to the task CPU is shown in table 4, the iteration number δ =1, and the initialization target value (the maximum number of the tasks carried by the unmanned aerial vehicle) O (δ -1) =0.
TABLE 1 parameter ranges for tasks
x(m) | y(m) | C(cycles) | D(bits) | |
|
0 | 0 | 1600000000 | 819200 |
Max | 100 | 100 | 16000000000 | 8192000 |
Table 2 parameter ranges for unmanned aerial vehicles
X(m) | Y(m) | H(m) | | |
Min | ||||
0 | 0 | 50 | π/6 | |
Max | 100 | 100 | 100 | π/2 |
TABLE 3 parameter Table for randomly generated tasks
x i (m) | y i (m) | C i (cycles) | D i (bits) | |
Task 1 | 34.9624 | 21.4531 | 333464936.6 | 3725299.33 |
Task 2 | 29.4299 | 7.2419 | 1373386716 | 4722915.82 |
Task 3 | 33.1190 | 40.7013 | 431874596.5 | 4982733.66 |
Task 4 | 61.0522 | 1.6279 | 1016263899 | 1614420.64 |
Task 5 | 90.6348 | 12.4428 | 1220885273 | 4668529.18 |
Task 6 | 4.6360 | 48.3225 | 1591534450 | 5971805.43 |
Task 7 | 85.7288 | 99.9832 | 1450304267 | 7445636.2 |
Task 8 | 13.9110 | 54.1779 | 603781186.8 | 4510664.35 |
Task 9 | 72.2903 | 6.4391 | 641641188.4 | 7501226.7 |
Task 10 | 91.5294 | 65.2428 | 878464385.9 | 3172692.16 |
Task 11 | 12.8824 | 0.8689 | 333843880 | 1578423.87 |
Task 12 | 60.2404 | 89.2471 | 1340989081 | 4741559.12 |
Task 13 | 15.3078 | 55.4631 | 806503241.9 | 883570.39 |
Task 14 | 21.3960 | 57.7343 | 864337501.5 | 7156238.06 |
Task 15 | 13.5097 | 37.2678 | 678734350.8 | 6822526.04 |
TABLE 4 initialize CPU computation frequencies assigned to tasks
Initializing CPU computation frequencies (cycles/bit) assigned to tasks | |
Task 1 | 444619915.5 |
Task 2 | 1831182288 |
Task 3 | 575832795.3 |
Task 4 | 1355018532 |
Task 5 | 1627847030 |
Task 6 | 2122045933 |
Task 7 | 1933739023 |
Task 8 | 805041582.3 |
Task 9 | 855521584.5 |
Task 10 | 1171285848 |
Task 11 | 445125173.4 |
Task 12 | 1787985441 |
Task 13 | 1075337656 |
Task 14 | 1152450002 |
Task 15 | 904979134.3 |
L2, solving the optimal position and elevation angle Q of the unmanned aerial vehicle and the transmission power P of the ground terminal equipment based on differential evolution and greedy algorithm:
and L2-1 converts the optimization problem P1 into an optimization problem P2 under the condition of given calculation frequency, and solves the optimization problem P2 by utilizing differential evolution and a greedy algorithm.
L2-2-1 constructs an initial population Q (τ) of P2 drones by equation (20) and table 2 and let τ =1.
TABLE 5 initial population of unmanned aerial vehicles Q (τ)
X(m) | Y(m) | H(m) | θ | |
j=1 | 71.02 | 75.39 | 52.35 | π/5.47 |
j=2 | 7.25 | 2.93 | 66.52 | π/5.88 |
j=3 | 81.40 | 0.43 | 68.47 | π/5.73 |
j=4 | 3.35 | 85.79 | 58.08 | π/5.59 |
L2-2-2 obtains the transmission power P (tau) of the ground terminal equipment through a critical boundary according to the formula (22) and the initial population Q (tau).
TABLE 6 Task i Transmission power p of i,j (τ)
L2-2-3 determines the unloading decision A (tau) according to the initial population Q (tau) and the table 7.
TABLE 7 Task i Is unloaded decision A (tau)
a i,1 | a i,2 | a i,3 | a i,4 | |
|
0 | 0 | 0 | 0 |
|
0 | 1 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 1 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 0 | 1 |
|
0 | 0 | 0 | 0 |
|
1 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
1 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
|
0 | 0 | 0 | 0 |
L2-2-4 finds the task target L =5 from the unload decision a (τ).
L2-2-5 was subjected to crossover and mutation operations according to the formulas (22) and (23) to obtain tables 8 and 9, which are shown in the following tablesAnd with
X(m) | Y(m) | H(m) | θ | |
j=1 | 77.89 | 75.00 | 60.87 | π/5.48 |
j=2 | 20.50 | 9.79 | 73.62 | π/5.85 |
j=3 | 74.53 | 0.81 | 59.95 | π/5.72 |
j=4 | 24.01 | 64.78 | 81.21 | π/6 |
X(m) | Y(m) | H(m) | θ | |
j=1 | 77.89 | 75.00 | 60.87 | π/5.48 |
j=2 | 20.50 | 9.79 | 73.62 | π/5.85 |
j=3 | 74.53 | 0.81 | 59.95 | π/5.71 |
j=4 | 24.01 | 64.78 | 58.08 | π/5.87 |
L2-2-6 from rand (0, 1) =0.7, U (τ) = Q can be obtained * (τ) for new ground terminal equipmentTransmission power P * (τ) and offload decision A * (τ) obtaining an evaluation L of the individuals * (τ)=8。
L2-2-7 reacting L (. Tau.) with L * Making a comparison if L (tau) < L * (τ), then A (τ + 1) = A * (τ)、P(τ+1)=P * (τ)、Q(τ+1)=Q * (τ) if L (τ) ≧ L * (τ), then a (τ + 1) = a (τ), P (τ + 1) = P (τ), Q (τ + 1) = Q (τ).
L2-2-8 judges whether tau is more than or equal to 1000, if yes, iteration is finished, and the optimal solution is obtained; if not, continue τ = τ +1 to L2-2-5.
L2-3 calculated the values of Q and P as shown in tables 10 and 11.
TABLE 10 position Q (τ) of drone by differential evolution max )
X(m) | Y(m) | H(m) | θ | |
j=1 | 77.89 | 75.00 | 60.87 | π/5.48 |
j=2 | 20.50 | 9.79 | 73.62 | π/5.85 |
j=3 | 74.53 | 0.81 | 59.95 | π/5.71 |
j=4 | 24.01 | 64.78 | 81.21 | π/5.87 |
TABLE 11 Task i Transmission power P (τ) max )
L3, solving an unloading decision and calculating a frequency scheme according to a convex optimization method:
l3-1 constructs a problem according to step S3-2-1 using a linearized approach to problem P3, constructing convex problem P6.
L3-2 calculates the solving convex problem P6 for each task using the KKT condition for each task separately according to step ii) in S3-2-2. Task based on KKT condition calculation 2 The computation is transmitted to the unmanned aerial vehicle for execution,ξ 1 =10 -2 、ξ 2 =10 -2 ,a 2,2 the results are shown in the tableShown at 12. Task computing completion Task i Offload decision a 1 As shown in Table 13, the relaxation variable γ i,j 、σ i,j 、ω i,j As shown in tables 14, 15, 16 and 17.
Table 12 Task uploaded to unmanned aerial vehicle 2
TABLE 13 Task i Offload decision a 1
a i,1 | a i,2 | a i,3 | a i,4 | |
Task 1 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 2 | 0.00 | 0.79 | 0.00 | 0.00 |
Task 3 | 0.37 | 0.00 | 0.00 | 0.00 |
Task 4 | 0.00 | 0.00 | 0.73 | 0.00 |
Task 5 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 6 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 7 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 8 | 0.00 | 0.00 | 0.00 | 0.45 |
Task 9 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 10 | 0.52 | 0.00 | 0.00 | 0.00 |
Task 11 | 0.00 | 0.43 | 0.00 | 0.00 |
Task 12 | 0.71 | 0.00 | 0.00 | 0.00 |
Task 13 | 0.00 | 0.00 | 0.00 | 0.49 |
Task 14 | 0.00 | 0.00 | 0.00 | 0.49 |
Task 15 | 0.00 | 0.44 | 0.00 | 0.00 |
TABLE 14 Task i Gamma of (2) i,j
TABLE 16 Task i Sigma of i,j
TABLE 17 Task i Omega of i,j
ω i,1 | ω i,2 | ω i,3 | ω i,4 | |
Task 1 | 0.00 | 4.2×10 -10 | 0.00 | 0.00 |
Task 2 | 6.2×10 -10 | 0.00 | 0.00 | 0.00 |
Task 3 | 0.00 | 0.00 | 5.56×10 -10 | 0.00 |
Task 4 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 5 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 6 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 7 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 8 | 0.00 | 0.00 | 0.00 | 5.27×10 -10 |
Task 9 | 0.00 | 0.00 | 0.00 | 0.00 |
Task 10 | 4.25×10 -10 | 0.00 | 0.00 | 0.00 |
Task 11 | 0.00 | 1.05×10 -9 | 0.00 | 0.00 |
Task 12 | 3.84×10 -10 | 0.00 | 0.00 | 0.00 |
Task 13 | 0.00 | 0.00 | 0.00 | 5.61×10 -10 |
Task 14 | 0.00 | 0.00 | 0.00 | 3.93×10 -10 |
Task 15 | 0.00 | 4.53×10 -10 | 0.00 | 0.00 |
The calculation frequency F of the unmanned aerial vehicle based on P and Q and allocated to the uploading task is obtained by the L3-3 according to the step ix) and the step x) in the S3-2-2 and is shown in a table 19, and the unloading decision A obtained by convex optimization is shown in a table 18.
And L3-4 obtains a final task unloading decision A according to S3-2-3, and obtains the calculation frequency F of the corresponding uploading task.
TABLE 18 Task i Offload decision a
TABLE 19 Task i Is calculated at a frequency F
L3-5 outputs the obtained unloading decisions A and F, and calculates a target value, namely the maximum number of the bearing tasks
L4-1 compares O (delta-1) =0 and O (delta) =9, if O (delta) ≧ O (delta-1), the O (delta) is kept unchanged, and the corresponding P, F, A, Q of O (delta) are recorded. If O (δ) < O (δ -1), O (δ) = O (δ -1), and the corresponding P, F, A, Q of O (δ -1) is recorded, let δ = δ +1.
L4-2 is equal to or greater than delta max And if not, stopping the loop iteration to output { P, F, A, Q, O (delta) }, obtaining the final { P, F, A, Q, O (delta) }, and if not, returning to the step S2.
Passing through delta max Obtaining O (delta) after the optimization of the second iteration max ) =15, the unmanned aerial vehicle position and elevation angle Q, the task offloading decision a, the calculated CPU frequency F, and the transmission power P of the ground terminal device are obtained at the same time, as shown in tables 20, 21, 22, and 22.
TABLE 20 offload decision A for task
a i,1 | a i,2 | a i,3 | a i,4 | |
|
1 | 0 | 0 | 0 |
|
0 | 0 | 0 | 1 |
|
1 | 0 | 0 | 0 |
|
0 | 1 | 0 | 0 |
|
0 | 1 | 0 | 0 |
|
1 | 0 | 0 | 0 |
|
0 | 0 | 1 | 0 |
|
1 | 0 | 0 | 0 |
|
0 | 1 | 0 | 0 |
|
0 | 0 | 1 | 0 |
|
0 | 0 | 0 | 1 |
|
1 | 0 | 0 | 0 |
|
1 | 0 | 0 | 0 |
|
1 | 0 | 0 | 0 |
|
1 | 0 | 0 | 0 |
TABLE 21 CPu calculation frequency F assigned to a task
TABLE 22 Transmission Power P of ground terminal equipments
Transmission power (W) | |
Task 1 | 0.025185832 |
Task 2 | 0.015680924 |
Task 3 | 0.051279243 |
Task 4 | 1.47 |
Task 5 | 0.014263526 |
Task 6 | 0.941542811 |
Task 7 | 1.5 |
Task 8 | 0.014619506 |
Task 9 | 1.49 |
Task 10 | 0.000301841 |
Task 11 | 1.48 |
Task 12 | 0.863828479 |
Task 13 | 0.00054 |
Task 14 | 1.5 |
Task 15 | 1.48 |
Table 23 optimal position Q of drone
X(m) | Y(m) | H(m) | θ | |
j=1 | 38.14 | 58.31 | 67.19 | π/5.36 |
j=2 | 74.61 | 18.54 | 59.32 | π/5.83 |
j=3 | 98.30 | 99.14 | 60.60 | π/5.69 |
j=4 | 0.55 | 2.81 | 50.31 | π/5.65 |
Claims (1)
1. The unmanned aerial vehicle deployment and task unloading method in the multi-unmanned aerial vehicle edge computing network comprises the following steps:
step 1: and constructing a mathematical model P1 of the ground terminal equipment task in the multi-unmanned-aerial-vehicle edge computing network.
Step 2: given the CPU calculation frequency F assigned to each task by the unmanned aerial vehicle, a mathematical model P2 is constructed. And solving a problem P2 by adopting a method of combining a differential evolution algorithm and a greedy algorithm, and solving the optimal position and elevation angle Q of the unmanned aerial vehicle, the transmission power P of the ground terminal equipment and the unloading decision A of the task.
And step 3: and (3) constructing a mathematical model P3 according to the optimal position and elevation angle Q of the unmanned aerial vehicle obtained in the step (2) and the transmission power P of the ground terminal equipment, completing non-convex-to-convex conversion by a linearization constraint factor method, constructing a mathematical model P6, solving and obtaining the CPU calculation frequency F and the unloading decision A which are distributed to tasks by the unmanned aerial vehicle through a convex optimization method, calculating the maximum bearing task number of the unmanned aerial vehicle, and marking the target value as L.
And 4, step 4: setting a maximum number of iterations δ max And =100, making the current iteration number delta =0, circularly iterating between the step 2 and the step 3, making delta = delta +1 every time iteration is carried out, comparing the maximized bearing task number of the unmanned aerial vehicle in the iteration, selecting the maximized bearing task number O (delta), and circularly iterating to delta max The final result converges and the maximum number of bearer tasks O (delta) is obtained max )。
Step 1, the steps of constructing a mathematical model of unmanned aerial vehicle deployment and task unloading in a multi-unmanned aerial vehicle edge computing network are as follows:
the Task of each ground terminal device i epsilon psi is expressed as a binary Task i =(C i ,D i ) In which C is i Representing processing tasks Task i The required CPU periodicity is expressed in cycles; d i Task representation i The unit is bit. (x) i ,y i ) Indicating the location, x, of the ground terminal equipment i e ψ i And y i The plane coordinates of the ground terminal equipment are all expressed in m, and the height is all 0 meter (m).
The 3-dimensional coordinates of each drone j ∈ φ are represented as (X) j ,Y j ,H j ) Wherein X is j And Y j Is the plane coordinate of the unmanned plane j epsilon phi, H j The height of the unmanned plane j epsilon phi is expressed, and the unit isAnd m is selected. And the unmanned plane j epsilon phi is provided with a directional antenna with adjustable beam width. The azimuth angle and elevation angle half-power beam widths of the unmanned aerial vehicle j epsilon phi antenna are equal and are all 2 theta j And e (0, pi) represents. Position and elevation angle Q of unmanned aerial vehicle j epsilon phi j =(X j ,Y j ,H j ,θ j ) The positions and elevation angles of all drones are denoted by Q.
a i,j Task i Unload decision of a i,j =0 for local calculation, a i,j =1 denotes Task to be executed i And (5) unloading to unmanned plane j for calculation. The unloading of the ground terminal device task comprises two stages: 1) Ground terminal equipment i epsilon psi sends Task i To drone j, 2) drone j allocates a computing resource computing task. Each ground terminal device is fixed in position before the mission is completed.
Task of S1-1 ground terminal equipment i epsilon psi i Execute locally
Task of ground terminal equipment i epsilon psi i The time of execution in the local is shown as (1):
whereinAnd the CPU frequency (in cycles/s) of the local processing task of the ground terminal equipment i epsilon psi is shown.
Ground terminal equipment i epsilon psi executing Task i The energy consumption calculated locally is shown in (2):
wherein eta l The effective capacitance coefficient represented by the ground terminal device i epsilon psi depends on the chip architecture of the CPU.
S1-2 Task i Offloading to drone execution
S1-2-1 communication model
The plane distance between the ground terminal device i epsilon psi and the unmanned aerial vehicle j epsilon phi is shown as (3):
unmanned plane j 1 With unmanned plane j 2 The distance between the planes is shown as (4):
unmanned plane j 1 With unmanned plane j 2 The distance between the two is shown as (5):
the azimuth angle half-power wave beam width and the elevation angle half-power wave beam width of the unmanned plane j epsilon phi antenna are equal, and the antenna gain modeling in the azimuth angle theta direction and the elevation angle u direction is shown in (6):
wherein G is 0 2.2846, ω =0 represents the channel gain outside the antenna beam width. For simplicity, ω =0 is set.
The user i belongs to psi and unloads the task to the unmanned plane j belongs to phi, and then the user i belongs to psiMust be in the coverage area of drone j ∈ φ, as shown in (7)
Channel gain g between ground terminal equipment i epsilon psi and unmanned aerial vehicle j epsilon phi i,j As shown in (8):
wherein beta is 0 Representing the channel power gain at a reference distance of 1 m.
The uplink transmission rate from the ground terminal device i epsilon psi to the drone j epsilon phi is shown as (9):
the uplink transmission power from the ground terminal device i e ψ to the drone j e φ is as shown in (10):
wherein p is i,j Representing the transmission power (in W) from user i e v to drone j e. B represents the channel bandwidth between the ground terminal equipment i epsilon phi and the unmanned aerial vehicle j epsilon phi, the bandwidth between all the ground terminal equipment and the unmanned aerial vehicle is equal, and the unit is Hz. N is a radical of 0 Representing the noise power spectral density (in W/Hz).
The transmission time from the ground terminal equipment i epsilon psi to the unmanned aerial vehicle j epsilon phi for unloading the task is shown as (11):
the transmission energy from the ground terminal equipment i e psi to the unmanned aerial vehicle j e phi for unloading the task is shown as (12):
s1-2-2 calculation model
Task i The time to unload to drone j e phi execution is shown as (13):
whereinIndicates that the unmanned plane j belongs to phi and is allocated to the Task i The CPU frequency (unit is cycles/s).
Task i The computational energy consumption offloaded to drone j e phi execution is shown in (14):
wherein mu c Representing the effective capacitance coefficient of the drone.
S1-3 problem description
Defining a mathematical model P1, and under the condition of meeting constraint conditions, maximizing the number of the bearing tasks of the unmanned aerial vehicle, wherein the model is as follows:
equation (15 a) is an objective function, χ = (F, a, Q, P) represents an optimization variable, whereRepresenting the CPU frequency allocation of the drone,the decision to offload a task is made, indicating the position and elevation of the drone,representing the transmission power of the ground terminal equipment.
Step 2, solving the unloading decision of the task, the optimal position and elevation angle of the unmanned aerial vehicle and the transmission power of the ground terminal equipment based on a differential evolution algorithm and a greedy algorithm, and the steps are as follows:
s2-1, constructing a mathematical model under the condition that the CPU of the unmanned aerial vehicle is assigned to each task to calculate the frequency F. And under the condition that the CPU calculation frequency F distributed to each task by the unmanned aerial vehicle is given, the unloading decision of the task, the optimal position and the elevation angle of the unmanned aerial vehicle and the transmission power of the ground terminal equipment are optimized, and a differential evolution algorithm and a greedy algorithm are adopted for solving. The solution target is consistent with the problem P1, and the number of the bearing tasks of the unmanned aerial vehicle is maximized. The mathematical model P2 of the optimization problem can be written as follows:
(16a) For the objective function, (16 b) - (16 k) are constraints, Ω = (a, Q, P) where Are the optimization variables.
S2-2, adopting a differential evolution algorithm and a greedy algorithm to solve unloading decision of tasks, optimal position and elevation angle of the unmanned aerial vehicle and transmission power of ground terminal equipment, and specifically comprising the following steps:
the problem P2 is solved, given the CPU calculation frequency F assigned to each task by the drone,firstly, solving the position and the elevation angle Q of the unmanned aerial vehicle based on a differential evolution algorithm, solving the transmission power P of ground terminal equipment and the unloading decision A of tasks based on a greedy algorithm, and finally obtaining the maximum bearing task number of the unmanned aerial vehicleSetting maximum iteration times tau of differential evolution algorithm max =10000, and let the current iteration number τ =1.
The steps of solving the problem P2 by the differential evolution algorithm are as follows:
i) The population is initialized as follows:
a coding mechanism is proposed in which a population set representsWhereinRepresents the set of the position and elevation of the Tth iteration unmanned aerial vehicle, Q j (τ)=(X j (τ),Y j (τ),H j (τ),θ j (τ)) represents the position and elevation coding for drone j ∈ φ, while noting Q j (τ)=q j (X j (τ),Y j (τ),H j (τ),θ j (τ)) represents the position and elevation angle of the drone j e Φ at the τ th iteration, and the value range is as follows:
wherein c is 1 And 0 each represents X j Upper and lower bounds of c 1 Representing a constant in meters (m). c. C 2 And 0 each represents Y j Upper and lower bounds of c 2 Denotes a constant with units of meter (m), H max And H min Each represents H j Upper and lower bounds in meters (m), θ max And theta min Respectively represent theta j An upper bound and a lower bound.
For convenience of description X j (τ),Y j (τ),H j (τ),θ j (τ) value range, let us assumeThe s decision variable representing the unmanned plane j belongs to phi, the value of s is {1,2,3,4 }, which respectively corresponds to Q j X in (τ) j (τ),Y j (τ),H j (τ),θ j (τ) that is
(17a) - (17 d) can be collectively expressed as follows:
Randomly generating decision variablesX j (τ),Y j (τ),H j (τ),θ j The initial population of (τ) is calculated as follows:
wherein rand (0, 1) denotes a random selection of a constant from 0-1.
The specific process of initializing the population comprises the following steps:
(1) let j =1,j denote the serial number of the drone, and randomly generate the position and elevation Q of drone j based on equation (20) j (τ)=(X j (τ),Y j (τ),H j (τ),θ j (τ)), and Q is added j (τ) put into the set of positions and elevations Q (τ) of drones for the τ -th iteration.
(2) Let iota =0, iota denote the generation of the j position and elevation Q of the drone j Number of formation of (. Tau.),. Iota max =200 denotes randomly generating drone j position and elevation Q j (τ) maximum number of generations.
(3) Randomly generating position and elevation angle Q of unmanned aerial vehicle j +1 j+1 (τ), iota = iota +1. Judging whether the position of each unmanned aerial vehicle in the unmanned aerial vehicle j +1 and the unmanned aerial vehicle position and elevation angle set Q (tau) of the tau iteration meets constraint expressions (16 g) and (16 j), namely Q (tau) and Q j+1 Whether or not the position of (τ) satisfies the constraint expressions (16 g) and (16 j). The judgment and selection are as follows:
if Q (τ) and Q j+1 (τ) satisfies the constraint equations (16 g) and (16 j), which indicates Q (τ) and Q j+1 (τ) distance between drones is greater than or equal to the nearest distance limitThat is, no collision occurs between drones in drones j +1 and Q (τ), and the coverage of drones in drones j +1 and Q (τ) is disjoint, then let j = j +1, and Q is assigned j (tau) putting the unmanned aerial vehicle position and elevation angle set Q (tau), and skipping to the step (4). If Q (τ) and Q j+1 (τ) does not satisfy constraint equation (16 f) and equation (16 j), indicating that the generation of the j +1 st drone position was unsuccessful. Iota is less than or equal to max And (4) if the answer is positive, skipping to the step (3) if the answer is positive, otherwise skipping to the step (1).
(4) And (3) judging whether j is less than or equal to N, and if yes, returning to the step (2). Otherwise, the positions and the elevation angles of all the unmanned planes are successfully generated, and the initial population Q (tau) of the unmanned planes is obtained.
ii) calculating the transmission power of the ground terminal equipment according to the new position and the elevation angle of the unmanned aerial vehicle, and the steps are as follows:
and according to the position and the elevation angle of the unmanned aerial vehicle j e phi and the position information of the user i e phi, the coverage range of which unmanned aerial vehicle the ground terminal equipment i belongs to is obtained through the formula (16 f). Calculating the frequency F and the maximum time delay T of the CPU assigned to a task for a given drone according to equation (16 b) max And obtaining the transmission rate required by the ground terminal equipment i epsilon psi to the unmanned aerial vehicle, wherein the transmission rate meets the following constraint:
as can be seen from the analysis formula (21), the transmission power of the ground terminal equipment must satisfy the formula (21), and meanwhile, the transmission power of the ground terminal equipment must also satisfy the formula (16 c), so that the equation (21) takes the equal sign as the optimal value of the transmission power of the ground terminal equipment.
Based on (21) and (10), the transmission power p of the tau iteration terrestrial terminal equipment i epsilon psi can be calculated i,j (τ) is represented by formula (22):
iii) Calculating the unloading decision of the task according to the transmission power of the ground terminal equipment and the new position and elevation angle of the unmanned aerial vehicle, and comprising the following steps:
dividing the task set psi into psi 1 、ψ 2 、ψ 3 Three categories, and satisfy ψ = ψ 1 ∪ψ 2 ∪ψ 3 。
Wherein the first type task psi 1 Having the highest priority, ground terminal equipment i epsilon psi 1 Indicating that it is not within the coverage of the drone, at which time Task i Can only be done locally, in which case the unloading decision a for the τ th iteration i,j (τ)=0。
Task psi of the second type 2 At this time, the ground terminal equipment i is epsilon psi 2 In the coverage of unmanned aerial vehicle, task i Can be unloaded to an unmanned aerial vehicle for execution, and the tau-th iteration ground terminal equipment i epsilon psi can be calculated based on the transmission power obtained in the step ii) 2 Transmission rate r to drone j i,j (τ) is calculated. Unmanned aerial vehicle assignment to Task i Under the condition that the calculation resources satisfy the formulas (16 b) and (16 c), the unloading decision is solved based on a greedy algorithm, and the selection strategy of the greedy algorithm is that the ground terminal equipment i belongs to psi 2 Transmission rate r to drone j i,j (τ), i.e. transmission rate r i,j (τ) fastest is selected first, with the unload decision a of the τ th iteration i,j (τ)=1。
Task type III 3 That is, after the unmanned aerial vehicle computing resources are fully loaded, the tasks can only be performed locally, namely, the total task psi is except psi 1 、ψ 2 Task outside, unload decision a for the τ th iteration at this time i,j (τ)=0。
iv) calculating a target value
Obtaining j e phi position and elevation angle of the unmanned aerial vehicle of the tau iteration based on the steps ii) and iii)The transmission power of the ground terminal equipment and the task unloading decision can obtain the result { P (tau), Q (tau) of the tau iterationAnd A (tau), wherein P (tau) represents the transmission power of the ground terminal equipment at the tau iteration, Q (tau) represents the unmanned aerial vehicle position set at the tau iteration, and A (tau) represents the task unloading decision at the tau iteration. Calculating a target value of the tau iteration based on the result of the tau iteration, namely the maximum number of bearing tasks of the unmanned aerial vehicle
v) after determining the position and the elevation angle of the tau-th iteration unmanned aerial vehicle, performing cross mutation operation, wherein the steps are as follows:
mutation operation: according toIs provided with For the unmanned plane j epsilon phi position and elevation angle obtained after the tau iteration variation, respectively correspond toBy mutation operations ofThe calculation formula of (a) is as follows:
equation (23) represents drone j for the τ th iterationEpsilon phi position and elevation angle q j (X j (τ),Y j (τ),H j (τ),θ j (τ)) performing a mutation operation, wherein j 1 、j 2 、j 3 E phi represents that 3 unmanned aerial vehicles with different numbers are randomly selected from the unmanned aerial vehicles, and tau represents Q in a differential evolution algorithm j F denotes the scaling factor, taking F =0.9.
And (3) cross operation: is also provided withRepresents the j ∈ phi position and elevation angle of the unmanned plane obtained after the tau-th iteration crossing according to Selected by cross probabilityWhereinThe calculation formula of (c) is as follows:
equation (24) represents that the position of the drone j ∈ phi at the τ -th iteration is crossed with the elevation angle, CR represents the cross probability, and CR =0.9.
Through mutation and crossing to obtainFinally obtaining a population U (tau) = { U = [) 1 (τ)、U 2 (τ)、…U j (τ)、…U N (τ) }. Steps ii) -iii) determining a new { P } * (τ)、Q * (τ)、A * (τ) }, step iv) obtaining the target value L * (τ)。
vi) the selection operation is as follows:
the selection is based on a greedy idea, i.e. the largest of the target values is selected as the new individual. Crossing the variation to obtain a population U (tau) = { U = [ (U) ] 1 (τ)、U 2 (τ)、…U j (τ)、…U N (τ) } is compared with the original population Q (τ) = { Q = 1 (τ)、Q 2 (τ)、…Q j (τ)、…Q N (τ) }, the comparison formula is shown in equation (25):
comparison L * (τ) and L (τ), as follows:
if L ≧ L (τ) indicates that the population obtained after the cross mutation operation is better, { P (τ + 1), Q (τ + 1), a (τ + 1) } = { P * (τ)、Q * (τ)、A * (τ) }, if L * (τ) < L (τ), then { P (τ + 1), Q (τ + 1), A (τ + 1) } = { P (τ), Q (τ), A (τ) } is unchanged.
vii) judgment of τ +1 > τ max And if yes, stopping loop iteration, and outputting { P (tau + 1), Q (tau + 1) and A (tau + 1) }, otherwise, tau = tau +1, and jumping to the step v).
Step 3, according to the position and the elevation angle Q of the unmanned aerial vehicle obtained in the step 2 and the transmission power P of the ground terminal equipment, solving a final unloading decision A of the unmanned aerial vehicle and a Task allocated to the unmanned aerial vehicle i The CPU calculates the frequency F, and the solving steps are as follows:
s3-1, solving an unloading decision and calculating resource allocation of the unmanned aerial vehicle, wherein the aim is to maximize the number of bearing tasks of the unmanned aerial vehicle, and a mathematical model P3 of an optimization problem can be written in the following form:
wherein x 1 = (A, F), whereinThe following equations (8), (9), (11) and (12) showAre all known values, i.e., are constants.
And S3-2, after the position and the elevation angle Q of the unmanned aerial vehicle and the transmission power P of the ground terminal equipment are obtained according to the step 2, an unloading decision A of the task and a computing resource allocation F of the unmanned aerial vehicle are solved by adopting a convex optimization method.
The specific operation of S3-2-1 non-convex-to-convex is as follows:
i) Order toAnd withRespectively representing the expressions (26 b) and (26 c), and obtaining the Hessian of (26 b) and (26 c)Matrix, as follows:
as can be seen from the analysis, since both equations (27 a) and (27 b) are non-positive definite matrices, (26 b) and (26 c) are non-convex constraints.
ii) carrying out a non-convex-to-convex transformation operation by a linearization constraint factor, comprising the following steps:
the equation (26 b) is first linearized, and in order to avoid the denominator being zero, a small variable epsilon is introduced 1 Equation (26 b) can be restated as:
an auxiliary variable gamma is defined i,j The method comprises the following steps:
substituting formula (29) into formula (28) to give formula (30):
substituting formula (29) into formula (26 e) to give formula (31):
since equation (26 e) represents the maximum CPU calculation frequency of the drone, it can be obtainedAlso, because of the formula (29), can obtainAnd replaceIn (1)The formula (32) can be obtained finally:
then model P3 may be rewritten to model P4 as follows:
since (33 b) and (33 c) are also non-convex, the equations (33 b) and (33 c) are relaxed as follows:
(1) for the relaxation operation of equation (33 b), the procedure is as follows:
due to the offload decision a i,j E (0, 1) is a discrete variable, thus relaxing a i,j E (0, 1), as shown in (34):
for formula (34), defineFrom equations (32) and (34), the constraint factor product ω can be obtained i,j :
(2) for the relaxation operation of formula (33 c), the procedure is as follows:
definition ofBecause of the fact thatTherefore, it is possible toThen σ can be obtained i,j The range of (c) is as follows:
model P4 can be transformed into P5 as follows:
(38),(39) (40g)
Because ofTherefore, it is possible toThe trend of (1) is the same, so the target formula is changed toA mathematical model P6 was obtained as follows:
S.t.(40b),(40c),(40d),(40e),(40f),(40g) (41b)
s3-2-2: adopting KKT conditions to solve the unloading decision of the unmanned aerial vehicle and the calculation frequency of the task, and specifically comprising the following steps:
i) Solving problem P6, constructing Lagrangian function and dual problem of P6, and orderingWherein A is 1 Represents the task unloading decision of the model P6 through convex optimization,lagrange multipliers of expressions (40 b) - (40 c),the Lagrangian multiplier of expression (40 d),the Lagrangian multiplier of expression (40 e),a Lagrangian multiplier of expression (40 f),each of the Lagrange multipliers represented by the formula (40 g),respectively, a Lagrange multiplier of formula (40 j), then
The lagrangian function of problem P6 is as follows:
ii) Lagrangian function (42) for each pairCalculating a partial derivative, wherein the steps are as follows:
lagrangian function (42) for a i,j The partial derivatives are calculated as follows:
iii) The conditions according to KKT are as follows:
from equation (43), equation (50) can be derived, as follows:
formula (51) is obtainable from formula (44):
formula (52) is obtained according to formula (45):
equations (53) and (54) are obtained from equations (46) and (47):
finding gamma in the formula (36) i,j 、ω i,j While the related equation can be found:
therefore, it can be analyzed from the formulas (55) and (56) that (53) and (54) are γ, respectively i,j Upper and lower bounds of (a), if γ i,j When the value between (52) and (53) does not take the boundary, thenAndare both 0.
Equations (57) and (58) are obtained from equations (48) and (49):
therefore, according to the equations (59) and (60), the values of σ of (57) and (58), respectively, can be analyzed i,j Upper and lower bound of (a), if i,j If the value between (57) and (58) does not take the boundary, thenAnd withAre both 0.
By analyzing (40 d), (40 e), (40 f) and (40 g), the results were obtained Since the portion is not 0, according to the KKT condition, are both 0. While not taking sigma i,j And gamma i,j The boundary of (1) is thenBoth are 0, initialization is performed so that i =1.
Equation (50) can be simplified to (61):
iv) initializationRespectively corresponding to convergence accuracy xi 1 、ξ 2 ρ is the number of iterations, let ρ =0, and the initial value is
v) obtaining a obtained from the rho-th iteration according to the formulas (51), (52) and (61) i,j (ρ)、ω i,j (ρ)、At this time The cases of (c) are as follows:
vi) updating Lagrange multiplier based on gradient descent method according to (44) and (45)As follows:
vii) judgment of a i,j (ρ)-a i,j (ρ -1) > 0, if yes, step v) is skipped, otherwise step viii) is skipped.
viii) judging whether i > M is satisfied, if not, i = i +1, turning to step iv), otherwise, outputting A 1 、ω i,j 、
ix) calculation task unmanned j assignmentTask i The calculation frequency of (c):
the Task can be obtained by the equations (29) and (48) i Is calculated frequency ofThe formula is as follows:
x) by A 1 The unload decision a is obtained as follows:
wherein A (a) i,j ) Representing a in the offload decision set A i,j ,A 1 (a i,j ) Representing offload decision set A 1 A in (a) i,j 。
by A 1 Reselecting an unloading decision A, starting to select j =1 from a first unmanned aerial vehicle, and specifically comprising the following steps:
i) Reselecting the unloading decision A through the residual calculation resources of the unmanned aerial vehicle, and comprising the following steps:
(1) selection A 1 0 < A in 1 (a i,j ) The value in the interval < 0.5, and secondly recording a in this interval i,j Task corresponding to the minimum value ofIs recorded as i 1 . Let A (a) i,j )=1,A 1 (a i,j ) And =0. Is re-paired by the formula (51)And (6) performing operation.
(2) If it isThenJudgment of 0 < A 1 (a i,j ) If < 0.5 is an empty set, the empty set jumps to step ii). Otherwise, returning to the step i).
first, A is selected 1 A in 0.5. Ltoreq 1 (a i,j ) The value in the interval of ≦ 1, and then recording a in this interval i,j Task of maximum value ofIs marked as i 2 。Order toIs newly paired by the formula (51)Carry out the operation ifThenIf it isAnd returning to the step i) again.
iii) Let j = j +1, when j > N, then obtain a final unloading decision a, and the CPU computation frequency allocated to the upload task by the drone is F.
iv) calculatingAnd O (delta) represents the number of the tasks carried by the unmanned aerial vehicle after delta iterations, and the values of the results F, A, Q and P of the optimization variables are obtained according to P2 and the formula (67).
Step 4 compares O (δ) with O (δ + 1), if O (δ 0+ 1) > O (δ 1), O (δ + 1) = O (δ + 1), and records F, a, Q, P for O (δ + 1), otherwise O (δ + 1) = O (δ), keeping F, a, Q, P for O (δ). Judging that delta +1 is larger than delta max And if not, making delta = delta +1, skipping to the step 2 for a new iteration based on the calculated F, and if so, ending the iteration to obtain F, A, Q and P of the optimized variables and the maximum bearing task number O (delta) of the unmanned aerial vehicle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211106470.7A CN115551013A (en) | 2022-09-12 | 2022-09-12 | Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211106470.7A CN115551013A (en) | 2022-09-12 | 2022-09-12 | Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115551013A true CN115551013A (en) | 2022-12-30 |
Family
ID=84725280
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211106470.7A Pending CN115551013A (en) | 2022-09-12 | 2022-09-12 | Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115551013A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116112981A (en) * | 2023-04-13 | 2023-05-12 | 东南大学 | Unmanned aerial vehicle task unloading method based on edge calculation |
-
2022
- 2022-09-12 CN CN202211106470.7A patent/CN115551013A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116112981A (en) * | 2023-04-13 | 2023-05-12 | 东南大学 | Unmanned aerial vehicle task unloading method based on edge calculation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108419286B (en) | 5G unmanned aerial vehicle communication combined beam and power distribution method | |
CN112134614B (en) | Downlink carrier resource allocation method and system for multi-beam communication satellite | |
Hashash et al. | Towards a decentralized metaverse: Synchronized orchestration of digital twins and sub-metaverses | |
CN111555795A (en) | Network resource allocation method and device based on user grouping scheduling | |
CN110856259A (en) | Resource allocation and offloading method for adaptive data block size in mobile edge computing environment | |
CN114268903B (en) | Geographic information assisted unmanned aerial vehicle relay position deployment and power distribution method | |
CN113613301B (en) | Air-ground integrated network intelligent switching method based on DQN | |
CN112672371B (en) | Air-ground collaborative hierarchical deployment model under heterogeneous demand and access method thereof | |
CN114880046B (en) | Low-orbit satellite edge computing and unloading method combining unloading decision and bandwidth allocation | |
CN115551013A (en) | Unmanned aerial vehicle deployment and task unloading method in multi-unmanned aerial vehicle edge computing network | |
Kisseleff et al. | A new optimization tool for mega-constellation design and its application to trunking systems [International Communications Satellite Systems Conference] | |
CN114584192A (en) | Anti-jitter robust beamforming optimization method for millimeter wave unmanned aerial vehicle communication system | |
Farajzadeh et al. | FLSTRA: Federated learning in stratosphere | |
CN112866368A (en) | Air-ground remote Internet of things design method and system | |
CN115664486A (en) | Energy efficiency optimization method for wireless energy supply in RIS (RIS) assisted UAV (unmanned aerial vehicle) edge computing system | |
CN113468819B (en) | Energy consumption optimization method based on genetic algorithm and adopting unmanned aerial vehicle to assist edge calculation | |
CN114599102A (en) | Method for unloading linear dependent tasks of edge computing network of unmanned aerial vehicle | |
CN113676917B (en) | Game theory-based energy consumption optimization method for unmanned aerial vehicle hierarchical mobile edge computing network | |
CN115767637A (en) | Cloud computing network resource optimal allocation method based on opportunistic access | |
CN115348558A (en) | Unmanned aerial vehicle deployment and calculation unloading joint optimization method based on convex optimization | |
CN114125744A (en) | Data acquisition method based on block chain rights and interests certification and terminal system | |
CN113342514A (en) | Edge calculation model based on near-earth orbit and service placement method thereof | |
CN117614520B (en) | Method for optimizing large-scale MIMO (multiple input multiple output) resources by removing cells based on unmanned aerial vehicle-satellite cooperation | |
Zhang et al. | Intelligent joint beamforming and distributed power control for UAV-assisted ultra-dense network: A hierarchical optimization approach | |
Hydher et al. | Recent Results of Machine Learning InspiredWireless Communications |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |